Scan candidate outreach, job descriptions, cover letters, and screening notes before they ship. Voice consistency across a five-plus TA team, sentence-level highlights to flag the AI passages, audit log on Business, REST API for ATS integration with Greenhouse, Lever, Workday, iCIMS, and Ashby. EEOC and ADA-aware framing built in. Free to try. No card.
In-house TA functions, embedded recruiting teams at growth-stage companies, RPO operations, contingent search firms, and the executive search side of any agency. Five to fifteen recruiters, two or three AI tools in the stack, fifty or more candidate touchpoints a day, and a hiring manager who reads the published outreach.
Recruiting teams answer to hiring managers who read the candidate-facing output. The CTO sees the engineering JD on Monday morning. The Head of Sales reads the InMail that went to a senior candidate. The CHRO opens the company careers page before a board update. Every candidate touchpoint gets read by someone with authority over the hiring function and no time to review drafts. The sent message is the only signal they have, and a templated AI tone reads as a brand problem rather than a recruiter problem.
Four to eight named recruiters on staff, often a senior recruiter or hiring lead, sometimes a recruiting operations specialist. Every recruiter now drafts with at least one AI tool. The variation between recruiters is healthy. The variation between AI tools layered on top of recruiters is the part that fragments employer voice across a quarter.
Contingent search, retained search, and RPO firms running candidate outreach at volume. Quality varies across recruiters and across the season. A shared scan standard with an Authenticity Score floor gives the firm one number to enforce instead of subjective voice notes that the junior recruiter cannot action.
The head of talent, the director of recruiting, or the VP of people owns the bar. Outreach review with the score in hand shrinks from ten minutes per InMail to ninety seconds. The recovered recruiter time pays for the workflow in the first week.
Four surfaces share one workflow: drafted with AI assistance, edited by the recruiter, reviewed by the hiring lead, scanned before sending. Each surface has its own register and its own risk profile when the AI flavour sneaks through.
The LinkedIn InMail or cold email that a senior candidate reads before opting in. Generic outreach signals a generic role, and reply rates on senior candidates collapse when the first line reads templated. Target an Authenticity Score above 80 on every outbound sequence and rewrite the flagged sentences before send. The opener and the call-to-action are the highest-risk lines because both default to stock phrasing when drafting moves fast.
Reply rates depend on conversational prose. A four-message follow-up drafted in one AI session reads as one voice across all four, and candidate replies collapse by message two. Scan the full sequence as a batch before scheduling and vary phrasing per message. Reply rates usually recover within a sequence or two once the scan becomes routine.
A JD written with AI can flag in candidate-side ATS scanners and browser extensions that warn applicants about templated postings. The senior candidate moves on without applying. Scan every JD before posting to Greenhouse, Lever, Workday, iCIMS, or Ashby, and rewrite the lines flagged at the sentence level. The hero paragraph and the requirements list are the surfaces where the templated tone shows up first.
Honest framing here. Detection on candidate writing is informational, not a rejection trigger. Scan the cover letter, look at sentence-level highlights, and use the result as a conversation starter in the phone screen rather than as a verdict. The next section covers EEOC and ADA implications in full.
Business at $39.99 a month standard, $29.99 a month on yearly, is the right fit for in-house talent acquisition teams running fifty or more candidate touchpoints a day across outreach, JDs, follow-up, and screening. Five shared seats, audit log, REST API, white-label PDFs. Full details on the pricing page.
Billed $89.88/year — Save $30
Billed $179.88/year — Save $60
Billed $359.88/year — Save $120
Yearly billing saves 25%. View full pricing →
AI detection is not a bias-free signal. Treating an Authenticity Score as a hiring gate can produce disparate impact on protected classes, and the EEOC has been explicit since 2024 that algorithmic screening tools fall under existing employment discrimination law. Calibrate carefully.
Detectors trained mostly on American English over-flag non-native English writing relative to native writing on identical-quality samples. Indian, Nigerian, Filipino, and other internationally-trained candidates writing formally taught English can score lower on the human axis even when they composed every word themselves. A hard rule like reject below 60 percent human will disproportionately affect candidates from these backgrounds.
Candidates with dyslexia, ADHD, autism, or other writing-related disabilities sometimes use AI tools as an accommodation to put their thinking into clean prose. A detection-driven rejection can effectively penalise a candidate for using an assistive tool, which raises ADA exposure. The candidate did not disclose the disability, and the recruiter rejected based on writing the candidate is legally entitled to assistance with.
The EEOC's May 2024 technical assistance document on AI in employment makes clear that an algorithmic tool used in selection is subject to the same disparate-impact analysis as any other selection device. If an AI detection threshold filters protected classes at a higher rate, the employer carries the burden of demonstrating job-relatedness and business necessity. A vendor-supplied detection score, on its own, is unlikely to clear that bar.
Use detection as a conversation starter in the interview, never as the sole reason to decline. Document the policy in writing before adopting any detector in the hiring funnel: detection results are informational input to the interview process, not a basis for screening or rejection. Train recruiters on the framing. Keep the audit log so the policy is visible if a hiring decision is ever challenged.
Senior candidates open three to seven recruiter messages a week and reply to two. The opener decides which two. Outreach drafted with ChatGPT and sent unedited reads templated on the first line, and the reply rate on senior candidates collapses.
Scan the first sentence and the second sentence as their own batch. The opener carries seventy percent of the open-to-reply conversion. If the opener scores low on the human axis, rewrite it before the rest of the message. Specific concrete detail about the candidate beats any phrasing trick the AI tool produces.
A four-message follow-up drafted in one AI session reads as one voice across all four. Scan the sequence as a batch, look at the sentence-level highlights, and rewrite the templated lines that repeat across messages. Reply rates lift noticeably on the second and third messages once the workflow is in place.
For senior roles, the hiring manager sometimes drafts the InMail and the recruiter sends it from their account. Run that draft through the scan before sending. A flat AI-flavoured message from a CTO reads worse than a clean message from a recruiter, because the candidate expects more from a hiring manager touch.
Most TA teams A/B test outreach copy at the subject line. Add the Authenticity Score as a second axis. Group A: opener above 85. Group B: opener above 90 with sentence-level highlights cleared. Track reply rate across both groups. The lift is usually visible within a hundred sends.
The candidate side of the market now runs AI detection on job postings. Browser extensions warn applicants about templated postings, and senior candidates skim and move on without applying. Pre-scanning JDs is the workflow change that keeps the role description distinct in a saturated market.
Tools like LazyApply, Simplify, and several Chrome extensions now scan job descriptions before the candidate applies and surface a templated-warning flag. Senior candidates with options use these tools to filter the inbound. A JD scoring low on the human axis loses applications from exactly the candidates the role is targeting.
Two surfaces inside a JD pull most of the AI-flavour weight. The hero paragraph that opens the posting, and the requirements list that follows. Both default to stock phrasing when drafted with AI. Scan the full JD, look at the highlights on these two surfaces specifically, and rewrite the lines flagged at the sentence level.
Inclusive-language tools that rewrite gendered or coded phrasing sometimes lift the AI score as a side effect, because the rewrites read as polished and templated. Run the inclusive-language pass first, then scan, then targeted human edits on the flagged lines. The two requirements collapse into one workflow.
Most TA teams paste the JD into the ATS as the last step before posting. Add a thirty-second scan before the paste. The Chrome extension scans inside the Google Docs or Notion draft where the JD lives. The Business tier REST API does the same scan automatically on a Greenhouse or Lever job-create webhook if you prefer it inside the ATS workflow.
Honest scope. The REST API on the Business tier covers any ATS via webhook. We do not ship a native plugin for any specific ATS yet, so integration is API-first with a few lines of glue. Most TA teams add the Chrome extension first and layer in the API once the workflow stabilises.
Webhook on job-create or candidate stage-change fires an API call to TextSight, returns the Authenticity Score and sentence flags as JSON, and writes the result to a custom field on the candidate or job record. Recruiters see the score inside Greenhouse or Lever without switching tools. The audit log lives in TextSight, the candidate record lives in the ATS, and the scan ID joins the two.
Enterprise ATS workflows on Workday and iCIMS use the same webhook pattern. Most TA teams running these systems already have a middleware layer (Workato, Zapier for Enterprise, or a custom integration) and the TextSight call sits inside that layer. The integration cost is half a day of work for a recruiting ops engineer or a third-party integration partner.
Ashby's webhook system is one of the cleanest in the ATS market and the TextSight integration is the simplest of the five. Most Ashby customers add the integration in an afternoon and have detection visible on the candidate record inside the week.
For the recruiters who spend most of their day inside LinkedIn Recruiter, the Chrome extension covers the lower-effort case. Scan an InMail draft from inside LinkedIn, scan a candidate cover letter from inside Gmail, and the score appears in a sidebar. No ATS integration needed for the per-message workflow.
Detection on candidate writing is informational, not a rejection trigger. Scan the cover letter, look at the sentence-level highlights, treat the result as a conversation starter in the phone screen. The line that matters is between AI as an aide and AI as the writer, and the interview is where you tell the difference.
A growing number of TA teams add a line to the application: "You may use AI as an aide. We will discuss your writing process in the interview." This removes the deception incentive entirely, because the candidate knows the topic is on the agenda. Application response quality goes up because candidates stop hiding the workflow they are using anyway.
Pick one or two flagged sentences from the cover letter and ask about them in the phone screen. "Your cover letter said the team accomplished X by doing Y. Can you walk me through how the team chose Y over the alternatives?" If the candidate composed the letter, they can answer. If the model composed it, the answer rarely lines up with the writing. Either way, the conversation surfaces the signal.
A single percentage is the wrong artifact for hiring. A cover letter scoring 78 percent AI tells you almost nothing on its own. Sentence-level highlights let you see whether the flagged passages are the analytical claims about the candidate's experience (which is closer to a problem) or the opening paragraph (which is closer to using AI for polish). The two cases call for different interview questions.
Read the EEOC and ADA section above before adopting any detection-based screening policy. The safe policy is detection as informational input to the interview, never as the sole reason to decline. The audit log on Business gives you a defensible record that the policy was applied consistently across candidates.
More for recruiters.
Sister B2B team page: brand voice across writers, agency coordination, audit log on Business.
For marketing →Team workspaces for search firms and RPO operations running fifty-plus candidate touches a day.
For agencies →REST API reference for ATS integration: Greenhouse, Lever, Workday, iCIMS, Ashby via webhook.
Read the docs →Free, Starter, Pro, Business. Yearly billing saves 25%. Business is the TA team tier.
See pricing →Free to try. No card. Business at $29.99 a month on yearly for in-house talent acquisition teams running fifty or more candidate touchpoints a day.